|
North American Network Operators Group Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical Re: A slight call to order (Re: Internic address allocation policy )
> > First, may I ask that when you reply to a message from the nanog mailing list, > you edit the headers so that they say "To: [email protected]" and have no CC? > Right now there is a strong penalty for anyone who adds to a thread, since we > will be on the CC list forever (getting two copies) even when it moves to a > different topic. We are all on the nanog list, no need to CC us. Ok :-) >Second, I've seen Karl and now Alan misuse a term. I'll pick on Alan since his >message is right in front of me, but the complaint is general (sorry Alan!): > > > Taking a relatively small chunk of the remaining address space > > (say, 210.*.*.*) gives us 64k addresses to hand out in convenient > > That's 16M addresses, not 64K addresses. We should not equivocate "addresses" > and "Class C networks". 210.*.*.* has 2^24 (minus subnet zero and broadcast > lossage) addresses -- 16M. 210.*.*.* has 2^16 "Class C networks" -- 64K. We > must not assume that every customer will get a Class C -- many will get just a > subnet since they will only have a handful of hosts. I know of several > providers who are chopping things up on nybble boundaries (16 hosts/net, or > actually 14 with the subnet zero and broadcast taken out). Not me! I consider a "Class B equivalent" to be 256 NETWORKS, by the common use of the term, but 65K *addresses*. For many of our account classifications we chop up network numbers to significantly smaller pieces. For some of them we chop them up to the *address* level, that is, ONE ADDRESS (1.2.3.4). For others we chop it up to the four-address level (mask 255.255.255.252), yielding three usable hosts if you can use "zero" (two if not) and for others we hand out larger blocks, including full Class "C"s. Right now, for political reasons, anything that a customer can take away from us is a class "C", primarily because not all providers are CIDR capable and the SWIP process is not really good for sub-C blocks, and until the backbone folks start insisting on this (we do for resale connections) this is the only way to insure that a customer doesn't get screwed somewhere down the line. A customer with a single SLIP connection gets one *address*; that is, 1.2.3.4, and that's it. Our backbone routes that internally. Yes, that is hell on the routers. However, it is also efficient on address space. -- -- Karl Denninger ([email protected])| MCSNet - The Finest Internet Connectivity Modem: [+1 312 248-0900] | (shell, PPP, SLIP, leased) in Chicagoland Voice: [+1 312 248-8649] | 6 POPs online through Chicago, all 28.8 Fax: [+1 312 248-9865] | Email to "[email protected]" for more information ISDN: Surf at Smokin' Speed | WWW: http://www.mcs.net, gopher: gopher.mcs.net
|